---
title: 'Open-RFNet: UAV RF Open-Set Recognition'
url: https://www.emergentmind.com/topics/open-rfnet
type: topic
---

# Open-RFNet: UAV RF Open-Set Recognition

Searching arXiv for the specified Open-RFNet paper and closely related context.
Open-RFNet is an open-set recognition model for UAV radio-frequency surveillance that was introduced in "Multi-Domain Supervised Contrastive Learning for UAV Radio-Frequency Open-Set Recognition" [2508.12689]. It is designed for the LA-ISAC setting enabled by 5G-Advanced, where unauthorized UAV activity creates a security problem for non-cooperative target monitoring. The system combines denoised RF preprocessing, dual-backbone feature extraction, multi-domain supervised contrastive learning, and an improved generative OpenMax procedure denoted IG-OpenMax. In the formulation reported by the paper, Open-RFNet targets both closed-set discrimination among known UAV types and open-set rejection of invasive UAVs that were not present during ordinary classifier training [2508.12689].

## 1. Problem setting and scope

Open-RFNet addresses UAV RF open-set recognition rather than conventional closed-set classification. In the reported experimental configuration, the dataset is DroneRFa, comprising 24 UAV types plus 1 "background"; the model uses 20 known classes and 5 held-out unknown classes, with RF collected at 915 MHz, 2.4 GHz, and 5.8 GHz by a USRP receiver [2508.12689]. The objective is not only to classify known UAVs but also to detect samples that should be assigned to an "unknown" category at inference time.

The pipeline begins from denoised 3 ms I/Q slices that are converted by STFT into a $775\times 775$ normalized power-spectrum matrix $\hat X[n,q]$ [2508.12689]. This preprocessing step is integral to the model definition in the paper, because denoising materially affects open-set behavior: the reported noisy-data versions attain $\mathrm{UAR}\sim 50$–$65\%$, whereas after denoising the reported UAR rises to $\sim 91$–$96\%$ [2508.12689].

A useful way to situate Open-RFNet is as an RF-domain architecture whose core challenge is unknown-class handling. This distinguishes it from the unrelated RGB-D semantic segmentation network RFNet for road-driving images [2002.10570] and from the ReFinement Network, also abbreviated RFNet, in point cloud completion [2112.03530]. The shared acronym does not denote a shared task family.

## 2. System architecture

Open-RFNet consists of three major blocks: preprocessing and feature extraction, closed-set training with supervised contrastive learning, and open-set recognition via IG-OpenMax [2508.12689]. The feature extractor is explicitly multi-stream.

The first stream extracts texture features through ResNet-18. The paper describes eight residual blocks, each following
$$
H_\ell = F_\ell(I_\ell;W_\ell) + I_\ell,\quad \ell=1,\dots,8,
$$
followed by global-average pooling and flattening to yield $\tilde z_a\in\mathbb R^{c}$ [2508.12689]. In parallel, two TransformerEncoder-based streams extract time-frequency position information. One branch operates in the time domain and the other on $\hat X^\top$ in the frequency domain; both use an initial small MLP, sinusoidal positional encoding, stacked TransformerEncoder layers, and a terminal small MLP to produce $\tilde z_b\in\mathbb R^d$ and $\tilde z_c\in\mathbb R^d$ respectively [2508.12689].

The three embeddings are then fused by concatenation and passed through another small MLP:
$$
\tilde z = \mathrm{MNLs}(\tilde z_a\oplus \tilde z_b\oplus \tilde z_c)\in\mathbb R^D.
$$
The fused representation is sent to a projection head for contrastive training and to a final classification head [2508.12689]. The paper characterizes this arrangement as a fusion of texture features and time-frequency position features from ResNet and TransformerEncoder.

| Component | Function | Output |
|---|---|---|
| ResNet-18 branch | Texture feature extraction | $\tilde z_a$ |
| TransformerEncoder time branch | Time-domain position modeling | $\tilde z_b$ |
| TransformerEncoder frequency branch | Frequency-domain position modeling | $\tilde z_c$ |

This architecture implies a deliberate separation between local texture-like RF signatures and positional structure in the time-frequency plane. A plausible implication is that the fused representation is intended to reduce failure modes associated with relying on only one RF characterization regime.

## 3. Multi-domain supervised contrastive learning

Closed-set training is organized around supervised contrastive shaping of the fused embedding space. After projection $\tilde z\mapsto z$, the paper minimizes the multi-domain supervised contrastive loss
$$
L_{\mathrm{sup}}
=\sum_{i\in I}
-\frac1{|P(i)|}
\sum_{p\in P(i)}
\log\frac{\exp\bigl(z_i\cdot z_p/\tau\bigr)}
{\sum_{a\in A(i)}\exp\bigl(z_i\cdot z_a/\tau\bigr)} ,
$$
where $I$ is the set of anchors in a batch, $P(i)$ denotes positives with the same true label as anchor $i$, $A(i)$ denotes the other samples, and $\tau$ is a temperature hyperparameter, reported as $0.07$ [2508.12689].

The training schedule reported in the paper is two-stage. First, the model undergoes SupCon pre-training for 30 epochs. Second, a linear softmax head is appended and the network is fine-tuned with cross-entropy for 10 more epochs [2508.12689]. The training details list batch size $128$ and Adam with cosine-annealing learning rate scheduling [2508.12689].

The paper attributes a specific representational role to the supervised contrastive objective: it pulls same-class features together and pushes different-class features apart, thereby balancing the ResNet texture stream and the Transformer-based position streams into a single embedding space with large inter-class margins [2508.12689]. Its ablation summary is correspondingly specific. Adding the Transformer alone or SupCon alone to ResNet decreases UAR, whereas only their joint use in Open-RFNet raises UAR from approximately $88\%$ to $96\%$ [2508.12689]. This suggests that the paper does not treat contrastive learning as a generic auxiliary loss, but as a mechanism that is effective only when paired with the multi-domain fusion design.

## 4. IG-OpenMax and open-set adaptation

The open-set component is an improved generative OpenMax algorithm, abbreviated IG-OpenMax [2508.12689]. The reported procedure is explicitly two-stage.

In the first stage, a conditional DCGAN trained with WGAN-GP is used on the known-class feature extractor outputs to generate "unknown" RF-feature samples. The WGAN-GP objective is given as
$$
L_{\mathrm{WGAN-GP}}
=\mathbb E_{x\sim p_r}[D(x)] - \mathbb E_{\tilde x\sim p_g}[D(\tilde x)]
+ \lambda\,\mathbb E_{\hat x\sim p_{\hat x}}
\Bigl(\|\nabla_{\hat x}D(\hat x)\|_2 - 1\Bigr)^2 .
$$
The paper then runs the closed-set model, denoted Open-RFNet-B, on GAN-generated samples and gathers the subset that are misclassified as simulated unknown-class features [2508.12689].

In the second stage, the entire feature extractor is frozen and only the final classification layer is retrained on the union of original known-class data labeled $1,\dots,N$ and the simulated unknown class labeled $N+1$ [2508.12689]. The paper identifies this as the crucial low-overhead step. It further states that the rationale for freezing is that simulating unknowns at the raw-signal or full-network level is hard and unstable; by operating in the final embedding space, IG-OpenMax needs only a small adjustment in the final classification layer [2508.12689].

At test time, Open-RFNet computes a mean activation vector $\mu_j$ for each class, fits a Weibull distribution $\mathrm{Weibull}(\tau_j,\kappa_j,\chi_j)$ to feature-to-MAV distances, recalibrates the top-$\alpha$ predicted classes with $\alpha=3$, and constructs an unknown logit from the residual mass before the final softmax [2508.12689]. The paper reports the recalibration weight for a top-ranked class $s(k)$ as
$$
c_{s(k)}(v)
=
1-\frac{\alpha-k}{\alpha}\,\exp\!\bigl(-(\|v-\mu_{s(k)}\|/\chi_{s(k)})^{\kappa_{s(k)}}\bigr),
$$
with the unknown dimension handled separately [2508.12689].

The resulting classifier therefore combines discriminative embedding formation with EVT-style post hoc score adjustment. A plausible implication is that the method uses contrastive training to make class manifolds more separable before asking OpenMax-style recalibration to carve an explicit unknown region.

## 5. Data pipeline, training protocol, and complexity

The preprocessing sequence begins with I/Q slicing to 3 ms segments, corresponding to 300 k points, followed by denoising by energy threshold, STFT to a $775\times 775$ magnitude spectrogram, and min-max normalization [2508.12689]. Known classes are used for SupCon pre-training and closed-set fine-tuning; unknown classes appear only at test time, except for GAN-simulated samples used during the adaptation phase [2508.12689].

The paper’s complexity analysis reports the following values for the full Open-RFNet and a plain ResNet-18 baseline:

| Model | Parameters | FLOPs per forward pass |
|---|---:|---:|
| Open-RFNet | $\approx 205.8$ M ($\sim 785$ MB) | $\approx 1.03 \times 10^9$ |
| Plain ResNet-18 only | $\approx 0.7$ M ($\sim 2.67$ MB) | $\approx 4.72 \times 10^8$ |

The same section states that adding the Transformer and fusion MLPs increases parameters by approximately $292\times$ and FLOPs by approximately $1.19\times$ relative to plain ResNet-18 [2508.12689]. This asymmetry is notable: parameter growth is dramatic, while the reported FLOP increase is moderate. The paper also notes that only the last linear classifier, now of size $(N+1)\times D$, is retrained during the open-set adaptation phase [2508.12689]. That design choice is central to the model’s claim of low-overhead open-set adaptation.

## 6. Empirical performance and comparative position

The paper reports multiple performance views, and they should be distinguished carefully. In the closed-set experiment on the 20-way known-class problem, Open-RFNet achieves $99.4\%$ closed-set accuracy [2508.12689]. In the open-set experiment with 5 held-out classes, the reported values are $\mathrm{KAR}=95.12\%$, $\mathrm{UAR}=96.08\%$, and $\mathrm{Gap}=0.96\%$ [2508.12689]. The abstract separately states that the model "achieves 95.12% in closed-set and 96.08% in open-set under 25 UAV types," so the paper presents both the 20-way closed-set accuracy figure and the KAR/UAR-based open-set view [2508.12689].

The open-set comparison table in the paper gives the following results:

| Method | KAR | UAR |
|---|---:|---:|
| Open-RFNet (IGOM) | 95.12% | 96.08% |
| OpenMax (Ours) | 95.23% | 93.84% |
| G-OpenMax (Ours) | 95.29% | 93.12% |
| S3R-OpenSet | 94.62% | 88.64% |
| UIOS-OpenSet | 93.98% | $\sim 0.00\%$ |

The same table reports $\mathrm{KP}=98.87\%$ and $\mathrm{UP}=83.86\%$ for Open-RFNet, while Open-RFNet-B and Open-RFNet-G exhibit lower UAR despite comparable KAR [2508.12689]. The paper therefore positions IG-OpenMax not as a large improvement in known-class retention, but as a stronger balancing mechanism between known and unknown recognition. That interpretation is supported by the reported gap values: $0.96\%$ for Open-RFNet versus $1.39\%$ for OpenMax, $2.17\%$ for G-OpenMax, and $5.98\%$ for S3R-OpenSet [2508.12689].

The t-SNE discussion in the paper further states that multi-domain SupCon disperses known classes and tightly isolates unknowns at the embedding level, enabling clean open-set boundary carving [2508.12689]. This suggests that the quantitative gains are linked to geometry of the learned representation rather than to recalibration alone.

## 7. Interpretation, limitations, and nomenclatural context

Open-RFNet is best understood as a composite system in which three elements are inseparable in the reported results: multi-domain feature fusion, supervised contrastive embedding shaping, and IG-OpenMax classifier adaptation [2508.12689]. The ablation summary argues against a common misconception that a stronger backbone or a contrastive objective alone is sufficient for RF open-set recognition; in the paper’s experiments, Transformer augmentation alone or SupCon alone decreases UAR, whereas their combination with IG-OpenMax yields the best open-set balance [2508.12689].

A second potential misconception is that open-set capability is learned directly from raw unknown samples. The reported method does not do this. Instead, it simulates unknowns in feature space with a conditional GAN, freezes the feature extractor, retrains only the classification head, and then applies Weibull-based recalibration at inference [2508.12689]. The model’s unknown-class handling is therefore partly synthetic and explicitly post hoc.

Within the broader literature represented here, the name "RFNet" is ambiguous. "Real-time Fusion Network for RGB-D Semantic Segmentation Incorporating Unexpected Obstacle Detection for Road-driving Images" defines RFNet as a two-branch encoder-decoder for RGB-D road scenes [2002.10570], while "A Conditional Point Diffusion-Refinement Paradigm for 3D Point Cloud Completion" uses RFNet to mean a ReFinement Network that predicts pointwise displacements after diffusion-based coarse completion [2112.03530]. Open-RFNet is unrelated to either architecture in modality and objective: it operates on UAV RF spectrograms, uses ResNet-18 plus TransformerEncoder branches, and targets open-set recognition rather than segmentation or geometric refinement [2508.12689].

Taken together, the reported evidence presents Open-RFNet as a UAV RF recognition framework whose primary contribution lies in embedding-space structuring for unknown rejection. Its empirical profile is characterized by $99.4\%$ closed-set accuracy on the 20-way known set, $\mathrm{KAR}=95.12\%$, $\mathrm{UAR}=96.08\%$, and a small reported known-unknown gap of $0.96\%$ in the open-set setting with 25 total classes [2508.12689].

Source: https://www.emergentmind.com/topics/open-rfnet